Pixel matching must be done because the coordinates of the object in different captured patterns are not one-to-one corresponding caused by the movement of object. A pixel matching method in on-line three-dimensional measurement based on phase prediction is proposed. Only one fixed sinusoidal fringe is needed to project on a measured object moving on the pipeline. While the object is coming to some positions with a certain adjacent displacement, the deformed patterns modulated by the object are captured synchronously with CCD camera. Dealing the captured deformed patterns with Fourier transform profilometry (FTP) method, the phase information of the object in those positions can be predicted to assist the pixel matching, which realizes the one-to-one corresponding of the object, and extracts the equivalent phase-shifting deformed patterns. So the 3D surface shape of the object moving on the pipeline can be reconstructed successfully with an equal phase-shifting algorithm. The results of simulation and experiment verified its feasibility and validity. Meanwhile, compared with the on-line FTP method, the root mean square error derived by the on-line FTP and the proposed method are 1.013 mm and 0.024 mm respectively, which shows the proposed method has higher precision in on-line three-dimensional measurement.